Mask-based diagnostic utilizing ai algorithms for improved patient outcomes

1 . A method of operating a diagnostic system using exhaled breath condensate (EBC) as a biosample, the diagnostic system comprising:

(a) an EBC collector having a condensate forming surface that converts exhaled breath vapor into an EBC liquid sample, the EBC collector being cooled before use so that the condensate forming surface is at a condensation forming temperature lower than a confined environment temperature inside of the face mask, and

(b) an EBC testing unit comprising (i) a biosensor configured to output an electrical signal, (ii) power, analysis and communications electronics in electrical communication with the biosensor, the analysis electronics configured to detect a target molecule dependent on the electrical signal received from the biosensor, and the communications electronics configured to communicate a detection of the target molecule, and (iii) a fluid conductor configured to conduct the EBC liquid sample to the biosensor; and the method comprising:

acquiring from an individual patient, using the biosensor, chemical biomarker data from the exhaled breath condensate (EBC) collected from the individual patient and non-chemical biometric data determined for the individual patient;

transmitting the chemical biomarker data and the non-chemical biometric data as individual patient data to a remote server;

aggregating the individual patient data with corresponding other patient data received from a set of other patients to form a first aggregated data set;

using the first aggregated data set to train a machine-learning AI agent to form a model to identify clusters having similar chemical biomarker and/or non-chemical biometric data as the individual patient and to estimate cluster-specific metrics, where the training includes

(i) applying an unsupervised algorithm to the first aggregated data set to assign the individual patient to a first order patient cluster that is a subset of the set of other patients,

(ii) constructing a labeled training data set from the chemical biomarker and/or non-chemical biometric data of the first order patient cluster, and

(iii) applying a trained model algorithm on the labeled training data set to compute first order cluster-specific metrics for the first order patient cluster and produce trained model parameters based on the labeled training data set, generating feedback, based on the trained model parameters, the feedback provided as automatically implementable actions to improve efficacy of the diagnostic system; and

providing the feedback to the diagnostic system for implementing the feedback to improve at least one hardware, software and networking component of the diagnostic system, including adjusting at least one operational parameter of the diagnostic system, the at least one operational parameter selected from the group consisting of a software setting, a hardware configuration, a network algorithm, a diagnostic threshold, a sampling frequency, a biomarker selection, and an alert routing.

2 . The method of claim 1 , where a software improvement of the diagnostic system operation is controlled based on the trained model parameters, where the software improvement includes automatically changing a user-interface to provide an indication to the individual patient to change at least one of a time of use and a frequency of use of at least one of a chemical biomarker testing system and a biometric detecting system used by the diagnostic system.

3 . The method of claim 1 , where a hardware improvement of the diagnostic system operation is automatically controlled based on the trained model parameters, where the hardware improvement includes changing a biosensor to detect a new chemical biomarker.

4 . The method of claim 1 , where a network improvement of the diagnostic system operation is automatically controlled based on the trained model parameters, where the network improvement includes changing algorithms to include at least one of a new AI agent, an algorithm to preprocess the individual data, the other patient data, the model to identify clusters, the unsupervised algorithm and/or the trained model algorithm.

5 . The method of claim 1 , further comprising aggregating a second aggregated data set comprised of the chemical biomarker and/or non-chemical biometric data from patients in the first order patient cluster, and applying the unsupervised algorithm, or another unsupervised algorithm, to the second aggregated data set; assign the individual patient to a second order patient cluster that is a subset of the first order patient cluster; constructing the labeled training data set from the chemical biomarker and/or non-chemical biometric data from patients in the second order patient cluster; and, then applying the trained model algorithm to compute second order cluster-specific metrics for the second order patient cluster to produce the trained model parameters.

6 . The method of claim 5 , where the individual patient data and the corresponding patient data include demographic information, and further comprising aggregating a third aggregated data set comprised of the demographic data of the second order patient cluster, constructing the labeled training data set from the demographic data and then applying the trained model algorithm to compute third order cluster-specific metrics for the third order patient cluster to produce the trained model parameters.

7 . The method of claim 1 , wherein training the machine-learning AI agent comprises executing a gradient-boosting procedure.

8 . The method of claim 7 where the gradient-boosting procedure:

(a) initializes a decision-tree model on collected and preprocessed patient data;

(b) iteratively identifies misclassified training instances, increases their weights, and fits additional decision trees focused on those instances; and

(c) combines the predictions of the decision trees with weights based on training performance to produce the trained model parameters,

the foregoing being performed by at least one processor and memory of the remote server.

9 . The method of claim 1 , wherein training the machine-learning AI agent comprises training a recurrent neural network (RNN).

10 . The method of claim 9 , wherein the RNN processes sequences of patient inputs comprising EBC biomarker measurements and non-chemical biometric measurements, the RNN maintaining an internal recurrent state across sequence elements to model temporal dependencies, and optimizing network parameters to reduce prediction error with respect to the labels of the labeled training data set, the foregoing executed by at least one processor and memory of the remote server.

11 . The method of claim 10 , wherein running the trained recurrent neural network on the at least one processor and memory of the remote server produces at least one of (i) a calibrated diagnostic risk score for the individual patient and (ii) trained model outputs specifying a control parameter vector; and further comprising generating, based on the trained model parameters, a control signal encoding the control parameter vector, and transmitting the control signal to the diagnostic system to automatically control the diagnostic system based on the trained model parameters, the control parameter vector specifying at least one of: a diagnostic decision threshold; an EBC collection sampling frequency or schedule; a selection of a biomarker panel or biosensor; and alert routing.

12 . The method of claim 1 , where the biosensor comprises at least one lateral flow assay and/or electronic biosensor.

13 . The method of claim 1 , wherein constructing the labeled training data set comprises, for each training example drawn from the first-order patient cluster: (a) forming an input from the individual patient's EBC chemical biomarker measurements and non-chemical biometric measurements for a defined window, (b) executing an Applied Probabilistic Analysis (APA) decision logic on the input using thresholded probability model outputs to assign a test-report class selected from Positive, Negative, or Maybe, and (c) storing the input together with the assigned test-report class as the label to create a labeled training example, optionally substituting a confirmed diagnostic outcome as the label when available.

14 . The method of claim 1 , further comprising preprocessing the chemical biomarker data and the non-chemical biometric data to form combined input variables for the training of the machine-learning AI agent, wherein the preprocessing comprises at least one of:

(i) time-aligning and/or resampling the biomarker and biometric measurements to a common cadence;

(ii) normalizing measurements relative to a baseline or control and suppressing outliers; and

(iii) performing principal component analysis (PCA) to reduce dimensionality;

and wherein the preprocessing is performed by at least one of the diagnostic system, a gateway device, and the remote server.

15 . A non-transitory computer-readable storage medium for operating a diagnostic system using exhaled breath condensate (EBC) as a biosample, the diagnostic system comprising:

(a) an EBC collector having a condensate forming surface that converts exhaled breath vapor into an EBC liquid sample, the EBC collector being cooled before use so that the condensate forming surface is at a condensation forming temperature lower than a confined environment temperature inside of the face mask, and

(b) an EBC testing unit comprising (i) a biosensor configured to output an electrical signal, (ii) power, analysis and communications electronics in electrical communication with the biosensor, the analysis electronics configured to detect a target molecule dependent on the electrical signal received from the biosensor, and the communications electronics configured to communicate a detection of the target molecule, and (iii) a fluid conductor configured to conduct the EBC liquid sample to the biosensor; and

the method comprising:

using the computer-readable storage medium including instructions that when executed by a computer, cause the computer to:

acquire using a first wearable electronic device worn by an individual patient, from the individual patient, using a biosensor, chemical biomarker data from the exhaled breath condensate (EBC) collected from the individual patient and non-chemical biometric data determined for the individual patient by the first wearable electronic device and/or a second wearable electronic device;

receive, at a remote server the chemical biomarker data and the non-chemical biometric data as individual patient data;

at the remote server aggregate the individual patient data with corresponding other patient data received from a set of other patients to form a first aggregated data set; and

using the first aggregated data set to train a machine-learning AI agent to form a model to identify clusters having similar chemical biomarker and/or non-chemical biometric data as the individual patient and to estimate cluster-specific metrics, where the training includes

(i) applying an unsupervised algorithm to the first aggregated data set to assign the individual patient to a first order patient cluster that is a subset of the set of other patients,

(ii) constructing a labeled training data set from the chemical biomarker and/or non-chemical biometric data of the first order patient cluster, and

(iii) applying a trained model algorithm on the labeled training data set to compute first order cluster-specific metrics for the first order patient cluster and produce trained model parameters based on the labeled training data set, generating feedback, based on the trained model parameters, the feedback provided as automatically implementable actions to improve efficacy of the diagnostic system; and

providing the feedback to the diagnostic system for implementing the feedback to improve at least one hardware, software and networking component of the diagnostic system, including adjusting at least one operational parameter of the diagnostic system, the at least one operational parameter selected from the group consisting of a software setting, a hardware configuration, a network algorithm, a diagnostic threshold, a sampling frequency, a biomarker selection, and an alert routing.

16 . The non-transitory computer-readable storage medium of claim 15 , where a software improvement of the diagnostic system operation is controlled based on the trained model parameters, where the software improvement includes changing a user-interface to provide an indication to the individual patient to change at least one of a time of use, a frequency of use, at least one of a chemical biomarker testing system and a non-biometrics detecting system used by the remote patient monitoring system.

17 . The non-transitory computer-readable storage medium of claim 15 , where a hardware improvement of the diagnostic operation is controlled based on the trained model parameters, where the hardware improvement includes changing a biosensor to detect a new chemical biomarker.

18 . The non-transitory computer-readable storage medium of claim 15 , where a network improvement of the diagnostic system operation is controlled based on the trained model parameters, where the network improvement includes changing algorithms running on the remote server to include at least one of a new AI agent, an algorithm to preprocess the individual data, the other patient data, the model to identify clusters, the unsupervised algorithm and/or the trained model algorithm.

19 . The non-transitory computer-readable storage medium of claim 15 , wherein training the machine-learning AI agent comprises at least one of training (i) a recurrent neural network (RNN), (ii) a gradient-boosted decision-tree ensemble, (iii) a decision tree, (iv) a random forest, or (v) a support vector machine.

20 . A diagnostic system comprising:

a first diagnostic system including (a) an EBC collector having a condensate forming surface that converts exhaled breath vapor into an EBC liquid sample, the EBC collector being cooled before use so that the condensate forming surface is at a condensation forming temperature lower than a confined environment temperature inside of the face mask, and

(b) an EBC testing unit comprising (i) a biosensor configured to output an electrical signal, (ii) power, analysis and communications electronics in electrical communication with the biosensor, the analysis electronics configured to detect a target molecule dependent on the electrical signal received from the biosensor, and the communications electronics configured to communicate a detection of the target molecule, and (iii) a fluid conductor configured to conduct the EBC liquid sample to the biosensor, and

a microprocessor configured to acquire from an individual patient chemical biomarker data from the biosensor and non-chemical biometric data determined for the individual patient, the communications electronics including a transmitter for transmitting the chemical biomarker data and the non-chemical biometric data as individual patient data;

a remote server for receiving the individual patient data for aggregation with corresponding other patient data received from a set of other patients;

an AI agent running on the remote server configured to analyze the aggregated data to identify patterns and trends and to dynamically control operation of the diagnostic device based on the identified patterns and trends, where the individual patient data is aggregated with the corresponding other patient data received to form a first aggregated data set, where the first aggregated data set is used to train the AI agent to form a model to identify clusters having similar chemical biomarker and/or non-chemical biometric data as the individual patient and to estimate cluster-specific metrics, where the training includes

(i) applying an unsupervised algorithm to the first aggregated data set to assign the individual patient to a first order patient cluster that is a subset of the set of other patients,

(ii) constructing a labeled training data set from the chemical biomarker and/or non-chemical biometric data of the first order patient cluster, and

(iii) applying a trained model algorithm on the labeled training data set to compute first order cluster-specific metrics for the first order patient cluster and produce trained model parameters based on the labeled training data set and generating feedback, based on the trained model parameters, the feedback provided as automatically implementable actions to improve efficacy of the diagnostic system; and

a receiver at the first diagnostic device for receiving the feedback from the remote server, wherein the microprocessor automatically reconfigures operation of the first diagnostic device in response to the feedback by modifying at least one physical or electronic operating parameter of the diagnostic device, including adjusting at least one operational parameter selected from the group consisting of a diagnostic threshold, an EBC sampling frequency, a biomarker detection configuration, and an alert routing parameter.

21 . The diagnostic system of claim 20 , wherein the AI agent is configured to personalize the selection of biomarkers to be detected based on a risk profile of the individual patient.

22 . The diagnostic system of claim 20 , wherein the AI agent is configured to control the first diagnostic device where the automatic control comprises at least one of:

dynamically adjusting thresholds for diagnostic test results;

modifying an applied therapeutic treatment;

adjusting the frequency of EBC collection and analysis;

updating a probability model for analyzing the aggregated data;

issuing alerts for emerging outbreaks or anomalies detected in population-level data;

issuing alerts to an individual patient or a caregiver;

modifying a software user-interface to suggest changes in time or frequency of use of a testing system;

changing a biosensor to detect a new biomarker;

changing algorithms running on a remote server to include a new AI agent;

creating a design of experiments (DOE) to optimize hardware or software components;

updating payer feedback or reporting systems based on improved patient outcomes or cost savings; and

refining testing thresholds based on applied probabilistic analysis.

23 . The diagnostic system of claim 20 , wherein the AI agent is configured to adjust user feedback to optimize the yield of exhaled breath condensate collection in real time.

24 . The diagnostic system of claim 10 , wherein the AI agent is configured to issue alerts to the patient based on regional public health data and detected outbreak trends.

25 . The diagnostic system of claim 20 , wherein the AI agent is trained and updated using clinical outcome data from a population of patients.

26 . The diagnostic system of claim 20 , wherein the non-chemical biometric data includes physiological signals obtained from wearable sensors.

27 . The diagnostic system of claim 10 , wherein the AI agent is configured to dynamically adjust the interval between sample collection events based on variation in biomarker levels.

28 . The diagnostic system of claim 20 , wherein the device is configured to provide diagnostic and trend data to a telemedicine provider dashboard.

29 . The diagnostic system of claim 20 , wherein the AI agent is configured to adapt the diagnostic operation for pediatric or adult patients.

30 . The diagnostic system of claim 20 , wherein the AI agent is configured to retrain its biomarker detection model based on newly received population-level data.

31 . The diagnostic system of claim 20 , further comprising a second diagnostic device for determining the non-chemical biometric data.

32 . The diagnostic system of claim 20 , where the one or more biosensors comprise at least one lateral flow assay and/or electronic biosensor.